mac table: a bundled address is named on the port-channel, not the member that carried it #1951

Merged
jaylouisw merged 1 commit from rev/1118-0701 into main 2026-11-18 07:30:40 +00:00
Owner

A real IOS switch learns a source address against the port-channel, not the member the frame arrived on. PacketForge learned on the member, so one address moved between Fa0/2 and Fa0/3 as the hash varied.

Evidence

  • Cisco, Catalyst 4500 command reference, show mac-address-table dynamic: "The output for an EtherChannel interface changes the port number designation (for example, 5/7) to a port group number (for example, Po80)" - the same sentence in five releases (12.2(20)EW, 12.2(25)EWA, 12.2(31)SGA, IOS XE 3.8/15.2(4)E).
  • Cisco, Validate Layer 2 Hardware on Catalyst 9000: 100 20bb.c05e.5318 DYNAMIC Po1 beside DYNAMIC Gi1/0/1, same table furniture this model prints.
  • Cisco Community thread (Apr 2024): a live Catalyst's DYNAMIC Po1 rows in three VLANs.
  • Cisco 6500 VSS note: the table "does not show ... which link in the etherchannel is selected for a particular packet" - the vendor draws exactly this distinction.

Change

  • macRows() names Po<id> for a member that is in the bundle; the physical port stays on the row as learnedOn, so show mac address-table interface <member> still answers.
  • A member that never negotiated (passive/passive) is named as itself - it forwards as an ordinary link, the same rule Simulator._logicalPortName gives the spanning tree.
  • One lag.portChannelName() shared with show etherchannel summary.

Unresolved (in research/mac-table-port-channel-name.md): show mac address-table interface Po1 does not resolve here (a bundle is not an interface object), and show spanning-tree still prints lag-1.

Gates: node scripts/syntax-check.js 407 files OK; full suite 334 suites, 11,570 assertions, 0 failed; new suite test/mac-table-port-channel-name.test.js 12 assertions.

A real IOS switch learns a source address against the **port-channel**, not the member the frame arrived on. PacketForge learned on the member, so one address moved between `Fa0/2` and `Fa0/3` as the hash varied. **Evidence** - Cisco, *Catalyst 4500 command reference*, `show mac-address-table dynamic`: "The output for an EtherChannel interface changes the port number designation (for example, 5/7) to a port group number (for example, Po80)" - the same sentence in five releases (12.2(20)EW, 12.2(25)EWA, 12.2(31)SGA, IOS XE 3.8/15.2(4)E). - Cisco, *Validate Layer 2 Hardware on Catalyst 9000*: `100 20bb.c05e.5318 DYNAMIC Po1` beside `DYNAMIC Gi1/0/1`, same table furniture this model prints. - Cisco Community thread (Apr 2024): a live Catalyst's `DYNAMIC Po1` rows in three VLANs. - Cisco 6500 VSS note: the table "does not show ... which link in the etherchannel is selected for a particular packet" - the vendor draws exactly this distinction. **Change** - `macRows()` names `Po<id>` for a member that is in the bundle; the physical port stays on the row as `learnedOn`, so `show mac address-table interface <member>` still answers. - A member that never negotiated (passive/passive) is named as itself - it forwards as an ordinary link, the same rule `Simulator._logicalPortName` gives the spanning tree. - One `lag.portChannelName()` shared with `show etherchannel summary`. **Unresolved** (in `research/mac-table-port-channel-name.md`): `show mac address-table interface Po1` does not resolve here (a bundle is not an interface object), and `show spanning-tree` still prints `lag-1`. **Gates**: `node scripts/syntax-check.js` 407 files OK; full suite 334 suites, 11,570 assertions, 0 failed; new suite `test/mac-table-port-channel-name.test.js` 12 assertions.
mac table: a bundled address is named on the port-channel, not the member that carried it
All checks were successful
CI / Classify the change (pull_request) Successful in 4s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 8s
CI / Server smoke test (pull_request) Successful in 6s
CI / Docker build (pull_request) Successful in 20s
CI / Test (Node 20) (pull_request) Successful in 50s
CI / Test (Node 22) (pull_request) Successful in 52s
CI / Test (Node 18) (pull_request) Successful in 57s
CI / Arm auto-merge (pull_request) Successful in 18s
54d6ef22de
`show mac address-table` learned on whichever bundle member the frame hashed to,
so the same source address was reported on Fa0/2 on one ping and Fa0/3 on the
next - a fact about one frame, printed where the reader expects a fact about
where the address is.

Cisco states the rule for this table: "The show mac-address-table dynamic
command output for an EtherChannel interface changes the port number
designation (for example, 5/7) to a port group number (for example, Po80)"
(Catalyst 4500 command reference, five releases), and a Catalyst 9000 table
prints `100 20bb.c05e.5318 DYNAMIC Po1` beside its physical-port rows.

- engine/device.js: macRows() names the bundle (`Po<id>`) for a member that is
  in the bundle, keeps the physical port on the row as `learnedOn`, and still
  names a member that never negotiated as itself, the rule the spanning tree
  already groups ports by.
- engine/lag.js: one `portChannelName()` so `show etherchannel summary` and the
  MAC table cannot call the same bundle two different things.
- engine/cli.js: `show mac address-table interface <member>` still finds what
  was learned on that cable.
- research/mac-table-port-channel-name.md: evidence, the two inferences, and
  what is unresolved (`interface Po1` does not resolve; `show spanning-tree`
  still says `lag-1`).
- test/mac-table-port-channel-name.test.js: 12 assertions.

Full suite: 334 suites, 11570 assertions, 0 failed. syntax-check: 407 files OK.
jaylouisw force-pushed rev/1118-0701 from 54d6ef22de
All checks were successful
CI / Classify the change (pull_request) Successful in 4s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 8s
CI / Server smoke test (pull_request) Successful in 6s
CI / Docker build (pull_request) Successful in 20s
CI / Test (Node 20) (pull_request) Successful in 50s
CI / Test (Node 22) (pull_request) Successful in 52s
CI / Test (Node 18) (pull_request) Successful in 57s
CI / Arm auto-merge (pull_request) Successful in 18s
to 985af11b26
All checks were successful
CI / Classify the change (pull_request) Successful in 7s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 6s
CI / Server smoke test (pull_request) Successful in 6s
CI / Docker build (pull_request) Successful in 20s
CI / Test (Node 20) (pull_request) Successful in 55s
CI / Test (Node 18) (pull_request) Successful in 1m2s
CI / Test (Node 22) (pull_request) Successful in 57s
CI / Arm auto-merge (pull_request) Successful in 7s
2026-11-18 07:28:45 +00:00
Compare
Sign in to join this conversation.
No reviewers
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
jaylouisw/packetforge!1951
No description provided.